GNN算法预测效果不好

[z12-zz]

PatchTST算法预测误差：
2025-11-07 11:21:18,011 - easytorch-training - INFO - Evaluate best model on test data for horizon 1, Test MAE: 0.0982, Test MSE: 0.0318
2025-11-07 11:21:18,011 - easytorch-training - INFO - Evaluate best model on test data for horizon 2, Test MAE: 0.1576, Test MSE: 0.0771
2025-11-07 11:21:18,017 - easytorch-training - INFO - Evaluate best model on test data for horizon 3, Test MAE: 0.2023, Test MSE: 0.1265
2025-11-07 11:21:18,017 - easytorch-training - INFO - Evaluate best model on test data for horizon 4, Test MAE: 0.2402, Test MSE: 0.1810
2025-11-07 11:21:18,017 - easytorch-training - INFO - Evaluate best model on test data for horizon 5, Test MAE: 0.2738, Test MSE: 0.2338
2025-11-07 11:21:18,020 - easytorch-training - INFO - Evaluate best model on test data for horizon 6, Test MAE: 0.2982, Test MSE: 0.2866
2025-11-07 11:21:18,020 - easytorch-training - INFO - Evaluate best model on test data for horizon 7, Test MAE: 0.3215, Test MSE: 0.3330
2025-11-07 11:21:18,020 - easytorch-training - INFO - Evaluate best model on test data for horizon 8, Test MAE: 0.3409, Test MSE: 0.3758
2025-11-07 11:21:18,020 - easytorch-training - INFO - Evaluate best model on test data for horizon 9, Test MAE: 0.3550, Test MSE: 0.4068
2025-11-07 11:21:18,020 - easytorch-training - INFO - Evaluate best model on test data for horizon 10, Test MAE: 0.3691, Test MSE: 0.4321
2025-11-07 11:21:18,020 - easytorch-training - INFO - Evaluate best model on test data for horizon 11, Test MAE: 0.3811, Test MSE: 0.4531
2025-11-07 11:21:18,020 - easytorch-training - INFO - Evaluate best model on test data for horizon 12, Test MAE: 0.3902, Test MSE: 0.4724
2025-11-07 11:21:18,020 - easytorch-training - INFO - Evaluate best model on test data for horizon 13, Test MAE: 0.3990, Test MSE: 0.4872
2025-11-07 11:21:18,020 - easytorch-training - INFO - Evaluate best model on test data for horizon 14, Test MAE: 0.4048, Test MSE: 0.5022
2025-11-07 11:21:18,020 - easytorch-training - INFO - Evaluate best model on test data for horizon 15, Test MAE: 0.4077, Test MSE: 0.5143
2025-11-07 11:21:18,020 - easytorch-training - INFO - Evaluate best model on test data for horizon 16, Test MAE: 0.4115, Test MSE: 0.5191
2025-11-07 11:21:18,020 - easytorch-training - INFO - Evaluate best model on test data for horizon 17, Test MAE: 0.4146, Test MSE: 0.5282
2025-11-07 11:21:18,020 - easytorch-training - INFO - Evaluate best model on test data for horizon 18, Test MAE: 0.4184, Test MSE: 0.5373
2025-11-07 11:21:18,028 - easytorch-training - INFO - Evaluate best model on test data for horizon 19, Test MAE: 0.4245, Test MSE: 0.5456
2025-11-07 11:21:18,028 - easytorch-training - INFO - Evaluate best model on test data for horizon 20, Test MAE: 0.4318, Test MSE: 0.5540
2025-11-07 11:21:18,028 - easytorch-training - INFO - Evaluate best model on test data for horizon 21, Test MAE: 0.4339, Test MSE: 0.5646
2025-11-07 11:21:18,031 - easytorch-training - INFO - Evaluate best model on test data for horizon 22, Test MAE: 0.4354, Test MSE: 0.5746
2025-11-07 11:21:18,031 - easytorch-training - INFO - Evaluate best model on test data for horizon 23, Test MAE: 0.4363, Test MSE: 0.5794
2025-11-07 11:21:18,031 - easytorch-training - INFO - Evaluate best model on test data for horizon 24, Test MAE: 0.4402, Test MSE: 0.5899
bigscity库的GNN预测结果：
MAE MAPE MSE RMSE masked_MAE masked_MAPE masked_MSE masked_RMSE R2 EVAR
1 0.245659 0.094156 0.145155 0.380992 0.245659 0.094156 0.145155 0.380992 0.971749 0.971838
2 0.407771 0.160088 0.374072 0.611614 0.407771 0.160088 0.374072 0.611614 0.927736 0.927940
3 0.522095 0.207343 0.617689 0.785932 0.522095 0.207343 0.617689 0.785932 0.881253 0.881483
4 0.611440 0.242147 0.869202 0.932310 0.611440 0.242147 0.869202 0.932310 0.833387 0.833636
5 0.691611 0.271292 1.123935 1.060158 0.691611 0.271292 1.123935 1.060158 0.785476 0.785720
6 0.763616 0.294849 1.373950 1.172156 0.763616 0.294849 1.373950 1.172156 0.740369 0.740619
7 0.821882 0.311618 1.595286 1.263046 0.821882 0.311618 1.595286 1.263046 0.702711 0.702894
8 0.862655 0.320675 1.776361 1.332802 0.862655 0.320675 1.776361 1.332802 0.670850 0.671015
9 0.893403 0.327805 1.941867 1.393509 0.893403 0.327805 1.941867 1.393509 0.636558 0.636778
10 0.922709 0.336573 2.086590 1.444503 0.922709 0.336573 2.086590 1.444503 0.606541 0.606787
11 0.955219 0.346685 2.226873 1.492271 0.955219 0.346685 2.226873 1.492271 0.580438 0.580722
12 0.982652 0.353163 2.358397 1.535707 0.982652 0.353163 2.358397 1.535707 0.552588 0.552923
13 0.995781 0.354742 2.449560 1.565107 0.995781 0.354742 2.449560 1.565107 0.533166 0.533599
14 1.005611 0.356003 2.515686 1.586091 1.005611 0.356003 2.515686 1.586091 0.518665 0.519198
15 1.016955 0.358762 2.567953 1.602483 1.016955 0.358762 2.567953 1.602483 0.506390 0.507041
16 1.026896 0.360662 2.603130 1.613422 1.026896 0.360662 2.603130 1.613422 0.498799 0.499566
17 1.031935 0.359173 2.617592 1.617897 1.031935 0.359173 2.617592 1.617897 0.496596 0.497396
18 1.033727 0.355199 2.630147 1.621773 1.033727 0.355199 2.630147 1.621773 0.502812 0.503739
18 1.033727 0.355199 2.630147 1.621773 1.033727 0.355199 2.630147 1.621773 0.502812 0.503739
19 1.026770 0.351446 2.620821 1.618895 1.026770 0.351446 2.620821 1.618895 0.504579 0.505324
20 1.023984 0.348893 2.634022 1.622967 1.023984 0.348893 2.634022 1.622967 0.502364 0.502939
21 1.028088 0.347460 2.654096 1.629140 1.028088 0.347460 2.654096 1.629140 0.502459 0.502973
22 1.030819 0.346181 2.665529 1.632645 1.030819 0.346181 2.665529 1.632645 0.501287 0.501668
23 1.031780 0.345927 2.693934 1.641321 1.031780 0.345927 2.693934 1.641321 0.494906 0.495192
24 1.037920 0.350122 2.738920 1.654968 1.037920 0.350122 2.738920 1.654968 0.486121 0.486435
请问GNN的MAE和MSE两个指标和其他算法对比效果差这么多，都使用了Standard归一化，是为什么呢